Loading your page…
Dermatology Practical & Conceptual
By Cansu Altınöz Güney, Huriye Aybüke Koç, Fadime Kılınç and 2 others.
Pityriasis versicolor (PV) is a common superficial fungal infection caused by Malassezia species. Potassium hydroxide (KOH) microscopy, the most widely used diagnostic method, has variable accuracy depending on observer expertise.
Ward brings these articles into the same app as your cards and questions. One free account for Ward, Suture, Osler and Prognia.
To develop convolutional neural network (CNN)-based deep learning (DL) models for automated PV detection from KOH images and compare their diagnostic performance with that of dermatologists.
Microscopic images (N = 1,921; 974 positive, 947 negative) from 20 KOH preparations of patients with suspected PV were evaluated. Using transfer learning, three pre-trained CNNs (MobileNetV2, ResNet50, VGG16) served as feature extractors with frozen convolutional backbones and custom classification heads. A 10-fold cross-validation (CV) strategy was applied. Performance was assessed using accuracy, sensitivity, specificity, positive predictive value, negative predictive value, F1 score, and area under the ROC curve (AUC). Additionally, 10 dermatologists independently evaluated 100 randomly selected test images.
In 10-fold CV, ResNet50 achieved the highest accuracy (93.2%) and F1 score (93.5%), while VGG16 had the highest AUC (0.953). MobileNetV2 showed the highest sensitivity (96.5%) and NPV (96.0%). On 100 test images, all models achieved higher accuracy than dermatologists (mean accuracy: 74.7%; sensitivity: 72.7%; F1 score: 0.714). Given dermatologists' lower sensitivity, MobileNetV2 appears most complementary for reducing false negatives in human-AI collaboration. Grad-CAM visualization from MobileNetV2 confirmed activation overlapping regions with characteristic Malassezia structures.
Deep learning models achieved higher accuracy than dermatologists in this single-image comparison. These findings support further investigation of DL models as clinical decision-support tools, requiring external prospective validation prior to implementation.
Published in Dermatology Practical & Conceptual. The abstract is shown as published and remains the work of its authors; read the full article at the source.
Prognia lists this article for reference and education. It is not medical advice. Always read the original and follow current guidelines and local protocols before acting on it.
